Across industries, the pressure to innovate within tight budgets is reshaping how technology leaders approach software development and delivery. Most are turning to AI as the obvious lever to drive speed and productivity. Nearly 9 in 10 IT leaders worldwide say they trust AI to make critical software release decisions – a figure that would have been unthinkable just a few years ago.
However, as AI adoption increases, a growing disconnect emerges between the rush to automate and the fundamentals of quality. Too often, businesses end up chasing speed at the expense of resilience. A recent global study of 2,700 technology practitioners and leaders found that nearly half openly prioritise delivery speed over software quality, and more than 60% admit to releasing code that hasn’t been fully tested, simply to meet deadlines.
It’s a trade-off with real consequences. Quality lapses erode customer trust, drive up maintenance costs, and increase the risk of security and compliance failures. In industries such as financial services or energy, the stakes are even higher; outages or critical bugs don’t just harm reputations; they can cost millions.
Why today’s AI agents fall short
Part of the problem lies in how most AI is currently deployed. Today’s AI agents are typically stateless; they’re designed to perform a single task, generate an output, and move on, without remembering past interactions or learning from historical patterns.
This makes for brittle systems. Every time an AI agent is asked to do something, it starts from scratch. There’s no accumulated context, no coordination with other AI processes, and no broader system memory to inform better decisions.
For enterprises, this often means layering narrow automation on top of fragile workflows, inadvertently increasing complexity rather than reducing it. While many leaders view AI as a means to meet tighter release cycles, the lack of context or shared intelligence can undermine long-term agility and reliability.
Enter Model Context Protocol: Building memory and collaboration into AI
That’s why there’s growing interest in Model Context Protocol, or MCP. Think of MCP as a kind of API for AI agents: a way to give them access to shared context and a persistent memory of past actions and outcomes.
Instead of working in silos, MCP-enabled agents can learn from organisational history and interact with each other to coordinate decisions. In practice, that might mean an autonomous testing agent that not only checks new code but also understands defect trends from previous releases and adjusts its coverage accordingly. Or deployment agents that time rollouts based on live insights from monitoring tools and historical risk patterns.
The result is a shift from isolated task automation to a more orchestrated intelligence. AI isn’t just speeding up single steps in the software delivery pipeline; it’s optimising the entire ecosystem, learning and adapting continuously.
The gap between ambition and readiness
Despite this promise, most organisations aren’t yet set up for an MCP-driven world.
Many teams are still grappling with fundamental coordination issues, including poor communication between developers and QA, disconnects between engineering leadership and delivery teams, and relentless pressure to push software out faster.
Meanwhile, new data highlights how ambition often outpaces readiness. While 82% of organisations are eager to offload repetitive tasks to agentic AI and free up time for more strategic work, nearly a quarter of technology leaders admit they lack full confidence in AI decision-making, a hesitation that’s especially pronounced in financial services and retail, where mistakes carry a steep cost.
In industries such as manufacturing and energy, over two-thirds face a significant risk of software outages within the next year, largely because quality processes struggle to keep pace with the rapid pace of innovation. For businesses already losing millions annually due to software quality failures, this gap underscores why building smarter, more resilient AI architectures is no longer optional.
How to start building for the future
None of this means businesses should pull back on AI. Quite the opposite: AI will only become more central to how organisations innovate and compete. But to unlock its full potential, the focus needs to expand from individual productivity gains to building architectures that support learning, context, and collaboration at scale.
That starts with asking a few tough questions: Where are our delivery processes most fragile today? Do our AI tools truly understand enough context to make consistently sound decisions? And are our data and quality governance practices mature enough to support more autonomous systems? These are not simple questions, but grappling with them is essential if we want to build the kind of resilient, future-ready operations that can thrive in an increasingly automated world.
Leaders also need to prepare teams. Upskilling shouldn’t just focus on model development or data pipelines but on designing and maintaining interconnected ecosystems where multiple AI agents can safely interact and improve together.
Finally, keeping an eye on emerging standards, such as MCP, will be crucial. The organisations that plan for interoperable, context-aware AI today will have a clear advantage as the technology and the competitive landscape continue to evolve.
A smarter way forward for AI-driven software
AI is no longer an experiment. It’s already helping businesses tackle demands they couldn’t otherwise meet, especially as budgets tighten and release cycles become more compressed. But chasing speed without the right foundations is risky. Without memory, context, and intelligent coordination, the promise of transformation can quickly collapse into costly technical debt and avoidable failures.
Model Context Protocol points to a smarter path forward: systems where AI agents don’t just automate isolated tasks but build on each other’s insights and learn over time. For enterprises looking to stay ahead, investing in these kinds of architectures isn’t just a future consideration; it’s fast becoming the dividing line between organisations that thrive and those left reacting to crises.
